Evaluating Conceptual Framing in AI-Generated Educational Writing: A Cognitive-Linguistic Analysis of Student Revision
Year: 2026
Type: preprint
Abstract: Large language models increasingly shape student writing, yet the conceptual assumptions embedded in their prose remain underexplored. This study examines how Gulf Arabic-speaking undergraduate writers revise the metaphorical representations of parental authority and child autonomy found in AI-generated argumentative essays. Drawing on Conceptual Metaphor Theory and Blending Theory, we compare AI-generated essays on parental influence over children's career choices with student-revised versions of the same texts. The analysis reveals the following revision patterns: students retain some metaphorical schemas, reduce or eliminate others, and introduce or expand alternative source domains with conceptually simpler and more student-centred alternatives. The analysis reveals the following revision patterns: students retain some metaphorical schemas, reduce or eliminate others, and introduce or expand alternative source domains. Most notably, the student revisions indicate that AI-generated conceptual framing is not simply carried over into the final texts; instead, parental authority is often reframed away from formative shaping and toward guidance, advice, and conditional involvement. The findings demonstrate how cognitive-linguistic analysis can reveal the conceptual assumptions embedded in AI-assisted writing and provide a framework for identifying potential cultural bias in educational uses of generative AI.
Source: Research Square
Authors: Marina Jovic, Mirka Ćirović
Institution: Gulf University for Science & Technology
Language: English
Subfield: Linguistics and Language
Field: Social Sciences
Domain: Social Sciences
SDG: Quality education
Open Access status: green
OpenAlex ID: W7167635441